06-19-2019 10:51 AM - edited 06-19-2019 11:11 AM
Solved! Go to Solution.
06-19-2019 11:26 AM
Dear Grothen,
The use of PARAM,AUTOMPC is very dangerous, I suggest to solve the potential conflict of "double-dependency" at user level and not let the problem in hands of the nastran solver, my suggestion is "never use PARAM,AUTOMPC,YES", the result will change?, of course!!.
If you need to define different RBE2 elemenets sharing the same dependent nodes but different location of the independent nodes, simple define groups with different configurations and activate the portion of the model you require for solving, and done!.
Best regards,
Blas.
06-19-2019 01:51 PM - edited 06-19-2019 01:52 PM
Blas,
Thanks for your reply.
Let's talk of simultaneous loads only in order to narrow the range of solutions.
If I understand correctly, you suggest creating an analysis and, in the Bulk Data Options, under "Portion of Model to Write", select a group which contains only 1 Rigid element (hence 1 load).
Repeat that step with different groups corresponding to the other rigid elements. Repeat as many times as I have rigid elements sharing same dependent nodes.
Then combine results with the Data Surface Editor.
Correct?
That's a solution and it will allow me to check if there is a difference in results when compared to using “PARAM,AUTOMPC,YES”.
However, do we understand the effect that swapping dependent/independent nodes can have on the linear static solver?
Thanks again,
06-21-2019 09:08 AM
I will leave the dependent/independent question to the NASTRAN experts. With regards to Blas' suggestion, he is suggesting multiple analysis sets, and in each one a reference to a group that contains your full model, and the one active rigid element, excluding the quasi-duplicates. You would create as many analysis sets as rigid/load configurations you have. Once you have the first group defined, you can copy it and flip in the next RBE. You will also need a load set for each configuration, and connect those into each analysis set when their node is included.
Mark.
06-21-2019 10:16 AM
I will respectfully disagree with Blas. In most cases, and the exceptions would be very unique, you should be fine with using AUTOMPC. The issue with independent/dependent dof's is strictly a book keeping issue in Nastran. Using this parameter will allow Nastran to rearrange the MPC equation that gets written internally, but the equation does not change, only the order of the terms in the equation. Take a look at documentation for MPC equation. Notice the first term of the input is identified as dependent, then followed by all of the other terms and coefficients. You can have the same equation, but use a different term for the first one, it is exactly the same relationship, therefor the model behaves the same, but a different node/dof is dependent.
Also, if you look at the RBE3 documentation, you will see that Nastran provides a "manual method" to rearrange this equation using the optional "UM" input. Unfortunately, there is no "manual option" for changing an RBE2, you must use AUTOMPC or change your model.
All that being said, using RBE2 to apply loads might not such a good choice since the equations that written internally result in no relative motion between the nodes that you connect. It is a true rigid connection and can result in a much stiffer model at those locations.
Traditionally, for the use case of spreading loads to nodes, the RBE3 is a better choice since it does not add any stiffness to the model. The equations that are written internally are very different for the RBE3.
Regards,
Joe
06-21-2019 05:35 PM
My best Kudos for Mark & Joe, thanks!.
Regards,
Blas.